RE: Keyboard regression by intel-vbtn

From: Limonciello, Mario
Date: Tue Sep 29 2020 - 16:38:41 EST


> > > >> I'm afraid that the only answer which I have to these questions
> > > >> is not helpful, but in my experience it is true: "firmware sucks".
> > > >
> > > > So FWIW there is a Dell 2-in-1 that has been conflated into this same
> issue.
> > > > https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1822394
> > >
> > > That is what a somewhat old kernel (5.0.0) which I guess may
> > > lack your fix to check the chassis-type.
> > >
> > > Interesting that this actually is a 2-in-1 though.
> > >
> > > Also interesting that according to the reporter this was
> > > triggered by a BIOS update.
> > >
> > > If you by any chance can provide an acpidump with both the
> > > 1.2.0 and 1.4.0 BIOS versions that would be very interesting.
> >
> > I requested on the Ubuntu bug for someone to provide these.
> >
> > > > Something that is confusing to me is that on the Windows side all these
> > > > machines use the same Intel driver for this infrastructure no matter the
> > > > OEM.
> > > > So they can't possibly be putting in quirk specific stuff in the driver
> side
> > > > can they?
> > > >
> > > > It has to make you wonder if some baseline assumptions made in the
> > > > driver early on around tablet mode support are completely false.
> > >
> > > I'm not saying your wrong. If you can get Intel to provide
> > > us with some documentation, or Windows driver source code
> > > for this, then that would be great.
> > >
> > > AFAICT the Linux driver currently is entirely based on
> > > reverse engineering.
> >
> > That's correct it was originally reverse engineered.
> >
> > Andy,
> >
> > As there is no publicly available documentation, could you see if it would
> it be
> > possible to get someone internal to Intel to compare private documentation
> to the
> > driver to see if something basic is missing or done wrong?
>
> I'm afraid that's all was designed solely for Windows and all information is
> not available to us in any form (basically somebody has to ask for WOS driver
> sources, and funny that as a customer your has more power to get this than
> mere
> in-house Linux engineer as me).
>

I guess this comes back to why we have reverse engineered driver in the first
place.

Even in this suggested case, such documentation is shared to partners under NDA,
and would need permission to be used in open source.

As such an allowlist for this driver might be the best path forward given the
quirky behavior that we have seen.